Accurate forest projections require long-term wood decay experiments because plant trait effects change through time.

Whether global change will drive changing forests from net carbon (C) sinks to sources relates to how quickly deadwood decomposes. Because complete wood mineralization takes years, most experiments focus on how traits, environments and decomposer communities interact as wood decay begins. Few experiments last long enough to test whether drivers change with decay rates through time, with unknown consequences for scaling short-term results up to long-term forest ecosystem projections. Using a 7 year experiment that captured complete mineralization among 21 temperate tree species, we demonstrate that trait effects fade with advancing decay. However, wood density and vessel diameter, which may influence permeability, control how decay rates change through time. Denser wood loses mass more slowly at first but more quickly with advancing decay, which resolves ambiguity about the after-life consequences of this key plant functional trait by demonstrating that its effect on decay depends on experiment duration and sampling frequency. Only long-term data and a time-varying model yielded accurate predictions of both mass loss in a concurrent experiment and naturally recruited deadwood structure in a 32-year-old forest plot. Given the importance of forests in the carbon cycle, and the pivotal role for wood decay, accurate ecosystem projections are critical and they require experiments that go beyond enumerating potential mechanisms by identifying the temporal scale for their effects.